Dominic Orsini

AI agents are only as useful as the context they can access — and most enterprise knowledge is scattered across BI tools, semantic layers, and unstructured documents, disconnected from the data agents act on. This session introduces the open context layer, which builds and maintains an open, customer-owned context layer directly in your warehouse or lake. Because Fivetran sits between every source system and the destination, we're uniquely positioned to capture context as it changes and reconcile conflicting definitions across sources, via metadata connectors, AI-ready document parsing, a context inference engine, and an MCP server for agent access. We'll also introduce Agents Schema, our new open-source standard for defining agent context inside the warehouse, next to your data, governed by permissions you've already set. No proprietary format. No lock-in. Just context, where your data already lives.

Following its demerger from Unilever, The Magnum Ice Cream Company rebuilt its data ecosystem from the ground up. In this session, discover how Magnum built a lean, modern platform for real-time analytics and AI by unifying data in a Databricks Lakehouse. Using Fivetran to integrate SAP, Salesforce, and Workday, the team reduced complexity and costs while shifting focus from pipeline maintenance to high-impact insights and AI innovation.